High levels of glycosylated human hemoglobin impair nitric oxide-mediated responses. However, the percentage of glycosylation for which this effect is observed and the mechanisms involved are unknown. We tested endothelium-dependent relaxations caused by acetylcholine in rat aortic segments either in control conditions or after preincubation with increasing percentages of glycosylated human hemoglobin. Human hemoglobin (1 and 10 nmol/L) inhibited endothelium-dependent relaxations only when glycosylated at 9% or higher. We evaluated the effect of 14% glycosylated human hemoglobin on acetylcholine-evoked responses in vessels preincubated with scavengers of superoxide anions, hydroxyl radical, or hydrogen peroxide (superoxide dismutase, deferoxamine, and catalase, respectively); with inhibitors of xanthine oxidase, cyclooxygenase, or thromboxane synthase (allopurinol, indomethacin, and dazoxiben, respectively); with blockers of thromboxane A2/prostaglandin H2 or endothelin receptors (SQ 30741 and BQ-123); and with the precursor of nitric oxide synthesis L-arginine. Superoxide dismutase abolished the effect of glycosylated hemoglobin, and the other substances did not have any effect. Glycosylated hemoglobin at 14% did not modify either the vasoconstrictions induced by the blocker of nitric oxide synthase NG-nitro-L-arginine methyl ester or the relaxations evoked in deendothelialized vessels by sodium nitroprusside and 8-bromo-cGMP. However, it inhibited the vasodilations evoked by exogenous nitric oxide. Superoxide dismutase abolished this latter effect. We conclude that the threshold for glycosylated human hemoglobin (Hb A1) to inhibit endothelium-dependent relaxation is 9%. This effect is due to interference with endothelial nitric oxide by means of superoxide anion production.